If anyone wants to perform normalization using a method not currently covered in Factor List, use Scaling Z-Score, N/A Handling None, Trim % 0, and Outliers Retain. This will yield the the original distribution scaled to standard normal allowing one to do any sort of transformation.
The mean and standard deviation can also be preserved in subsequent downloads by using a fixed ML Training End date. (While this is generally true, any change in the universe or data would affect the result.) This unfortunately has a growing API cost since one must download the same data multiple times, but it’s the solution we currently have.
Some of this was mentioned here: No need for raw scores - #4 by marco